build: add commit hooks

.githooks/check-large-files.sh

@@ -0,0 +1,14 @@
+#!/bin/sh
+# Block staged files over the limit: checkpoints, datasets and binaries stay out of git.
+# Called by lefthook as: check-large-files.sh {staged_files}
+limit_mb="${MAX_FILE_MB:-5}"
+limit=$((limit_mb * 1024 * 1024))
+status=0
+for file in "$@"; do
+  size=$(git cat-file -s ":$file" 2>/dev/null) || continue
+  if [ "$size" -gt "$limit" ]; then
+    echo "large file: $file is $((size / 1024 / 1024)) MB (limit ${limit_mb} MB). Add it to .gitignore." >&2
+    status=1
+  fi
+done
+exit $status

.githooks/commit-msg-check.sh

@@ -0,0 +1,27 @@
+#!/bin/sh
+# Conventional Commits with a short subject; no AI attribution.
+# Called by lefthook as: commit-msg-check.sh <message-file>
+msg_file="$1"
+subject=$(sed -n '1p' "$msg_file")
+
+case "$subject" in
+  Merge\ * | Revert\ * | fixup!\ * | squash!\ * | amend!\ *) exit 0 ;;
+esac
+
+types='feat|fix|refactor|perf|docs|test|build|ci|chore|style|revert|exp'
+if ! printf '%s\n' "$subject" | grep -Eq "^($types)(\([a-z0-9._/-]+\))?!?: [^ ].*$"; then
+  echo "commit-msg: subject must be '<type>(<scope>): <summary>'" >&2
+  echo "  types: ${types}" >&2
+  echo "  got:   $subject" >&2
+  exit 1
+fi
+
+if [ "${#subject}" -gt 72 ]; then
+  echo "commit-msg: subject is ${#subject} chars, keep it at 72 or less (aim for ~50)" >&2
+  exit 1
+fi
+
+if grep -Eiq 'co-authored-by:[[:space:]]*claude|noreply@anthropic\.com|generated with \[?claude code' "$msg_file"; then
+  echo "commit-msg: no AI attribution lines in commit messages" >&2
+  exit 1
+fi